| Summary: | Introduction and Objectives: Routinely recorded clinical data held in administrative healthcare databases have demonstrated utility in Randomised Controlled Trials (RCTs), such as informing recruitment feasibility assessments, assisting with recruitment and measuring certain study outcomes. Furthermore, routinely recorded non-clinical data, such as data regarding employment and taxation have the potential to inform the measurement of outcomes including health economic analyses. However, limitations with accuracy, access and feasibility have been identified. This research systematically reviewed the use of routinely recorded data in the UK in RCTs and agreement of routinely recorded data compared to data collected using standard prospective research methods. Subsequently, the accessibility, quality, agreement and feasibility of using routinely recorded data compared to data collected using standard prospective methods: in a UK RCT assessing antiepileptic drug treatments for individuals newly diagnosed with epilepsy were assessed. Methods A systematic review was undertaken to assess the use of routinely recorded data in the UK in RCTs and to compare agreement between routinely recorded data and data collected using standard prospective methods. The quality of routinely recorded data was assessed according to a number of criteria, such as the degree of missing data. The agreement of routinely recorded data compared to data collected using standard methods in a UK RCT included calculation of Cohen's Kappa and construction of Bland Altman Plots for categorical and continuous variables, respectively. Assessment of the resources required for protocol development, application for data, accessibility, data preparation and attributes including data coverage and agreement were included in the assessment of feasibility. Results: The reviews identified few published studies using routinely recorded clinical data in the UK in RCTs. Mortality data were most commonly used and demonstrated the greatest agreement compared to data collected using standard methods. There was no evidence for the use of routinely recorded data from non-clinical sources, such as governmental employment and taxation data in clinical RCTs. There was a significant degree of missing routinely recorded data identified for variables and outcome measures relevant to the case study RCT, compared to data collected using standard methods. Where paired data were available, agreement was not satisfactory for the majority of comparisons, including the assessments of seizure occurrence and adverse events. Agreement was satisfactory for two comparisons; date of prescription of antiepileptic drugs and healthcare resource use, assessed using the dates of RCT follow-up assessments. The poor accessibility, prolonged period of application and cost for data access together with the poor data quality and agreement resulted in the limited feasibility for use of routinely recorded data in this RCT assessing antiepileptic drug treatments for epilepsy. Conclusions: There is currently very limited experience of using routinely recorded data in UK RCTs for outcomes other than mortality and healthcare resource use for economic evaluations. In this research the accessibility and feasibility of use were limited and degree of missing data and agreement compared to data collected using standard methods unsatisfactory. The results of this research suggest routinely recorded data in the context of prospective clinical research could be an important source of additional data, for example to identify additional events such as seizures not recorded using standard methods. The results suggest that use of routinely recorded data as the primary data source or as a means of validating data collected using standard methods, would be limited. Recommendations include suggestions for improving the access to routinely recorded data for research, development of an integrated electronic health record for use in both clinical practice and research and further assessment of the attributes and 'optimal mix' of routinely recorded data compared to data collected using standard methods.
|
|---|
